Contigency table visual representation for notebooks

[arshiaar]

Please work through the following checklists. Delete anything that isn't relevant.

Development for new xarray-based metrics

• [x ] Typehints added

Docstrings

• Docstrings complete and follow Napoleon (google) style
• Code example added

Tutorial notebook

-[x] update tutorial notebook

[arshiaar]

#361

[tennlee]

Hi @arshiaar. We had discussed some adjustments to the dataframe labels. Can you please indicate if you plan to continue this PR to implement them? If not, that is completely fine, I will happily progress the work from here, but I wanted to give you the chance to let us know if you'd prefer to do this yourself.

[tennlee]

I have rebased this work (brought it up to date against the 'develop' branch) and pushed up that change into this PR.

[tennlee]

@nicholasloveday This PR is now at a review-ready state.

[tennlee]

@arshiaar I removed the _repr_html method. As a side note, it needs to be named repr_html to be recognised by JupyterLab. However, the function doesn't work in all cases. If any dimensionality has been preserved, the confusion matrix is no longer 2x2. As such, I thought it was perhaps not ideal to have as default rendering function. I'm not entirely sure how higher-dimensional confusion matrixes should be handled, so I just removed it as a default renderer and put a note in the docstring, which I think is sufficient. I think a lot of the use cases will still be met, and this is a nice representation of the most common 2x2 case. Thanks again for all your help.

[nicholasloveday]

This looks to work nicely for the 2x2 case.

For higher dimensions, people can do something like

new_manager = contingency_manager.transform(preserve_dims='lead_hour')
new_manager.get_table()

It doesn't display a nice table, but it people can access the data and make their own tables.

One thing that would be nice to do is to provide a more useful error message if you call format_table on a higher dimensional matrix

new_manager = contingency_manager.transform(preserve_dims='lead_hour')
new_manager.format_table()

[tennlee]

Thanks Nick. I had been wondering about whether to go to the extra effort to provide some more error-handling, and in light of your comments I will do so, and post here when I have something implemented.

[tennlee]

@nicholasloveday I have updated the PR with some additional handling code - what do you think?

[nicholasloveday]

Thanks - the update looks good to me!